Engineering Failure Analysis 47 (2015) 89–101

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

An investigation of premature fatigue failures of gas turbine

blade
Wassim Maktouf a,b, Kacem Saï b,⇑
a
Mechanical ATA, BG Tunisia
b
UGPMM, National Engineering School of Sfax, Tunisia

a r t i c l e i n f o a b s t r a c t

Article history: A failure of a first stage compressor blade of a Gas Turbine Generator in a Gas Treatment
Received 4 July 2014 plant caused severe mechanical damage to the compressor section and power supply
Received in revised form 26 September 2014 troubles. In this paper, the blade failure is investigated by mechanical, metallography
Accepted 30 September 2014
and chemical analysis. A finite element analysis is performed on the blade geometry to
Available online 20 October 2014
identify the stress concentration areas and the stress/strain values. The investigation
outcomes provided the most probable cause of the premature blade failure and the recom-
Keywords:
mendations to mitigate such incidents.
Turbine blade failures
Metallography and chemical analysis
Ó 2014 Elsevier Ltd. All rights reserved.
Finite element analysis

1. Introduction

Despite the new designed engines became more efficient and durable, blade failures often lead to loss of all downstream
stages and can have a dramatic effect on the availability of the turbine engines. In fact, the majority of gas turbo-generators
are used as an auxiliary compensator to generate electric power at the peak of load demand. Therefore they are often utilized
in discontinuous conditions of commissioning which leads to a lot of shocks, risks and premature blades’ failures.
Several GT (Gas Turbine) engineering handbooks and technical papers indicate that the conditions which influence the
blade lifetime are the high mechanical stresses due to centrifugal force, vibratory and flexural stresses; the operation envi-
ronment (high temperature, fuel and air contamination, solid particles, etc.) and the high thermal stresses due to thermal
gradients. However the degree of deterioration in an individual blade differs due to several factors such total service time
and operation history (number of start-ups, shut-down and trips), engine operational conditions (temperature, rotational
speed, mode of operation: base load, cyclic duty, etc.) and manufacturing differences (forging, casting, heat treatment, grain
size, porosity, alloy composition, etc.) [1,2].
Statistics on industrial GT [3] indicate that blade failure represents 62% of the total damage costs for heavy duty GT. High
cycle fatigue (HCF) occurs for 12% of compressor blades [2]. Thus, several technical papers have been interested in the fatigue
aspects of the GT blades. Kargarnejad and Djavanroodi [4] have performed an assessment of a failed GT blade and noticed
that the maximum stress due to centrifugal force and fluid pressure is located near the connection point of the airfoil and
the root. Crack initiation and propagation in the base metal was due to mixed fatigue/creep mechanism and grain boundary
brittleness caused by formation of a grain boundary continuous film of carbides. Qu et al. [5] have investigated the fracture
surface of a failed first stage blade in GT engine and found that during initial fracture stage, the crack propagates slowly and
the fracture surface is generally quite flat. Accordingly, the fracture surface becomes gradually rougher because of the

⇑ Corresponding author. Tel.: +216 74 274088; fax: +216 74 275595.

E-mail addresses: wassim.maktouf@bg-group.com (W. Maktouf), kacemsai@yahoo.fr (K. Saï).

http://dx.doi.org/10.1016/j.engfailanal.2014.09.015
1350-6307/Ó 2014 Elsevier Ltd. All rights reserved.
90 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101

increasing crack propagation rate at higher real stress and smaller carrying area. Fracture veins/striations and clam shell
markings of the fracture morphologies displays a fatigue fracture features and indicate the propagation direction of the
fatigue crack. On other hand dendrite morphology of the surface indicates instantaneous fracture features mainly due to last
stage of the fracture process or aroused by hitting the initial broken blade at high rotating speed.
Other studies focused on determining the vibrational characteristics of the first row blades by static and dynamic finite
element (FE) analysis [6–9]. The dynamic behavior of the blade was investigated using modal and harmonic analysis and the
drafted campbell diagram reveals the coincidence points between the rotor exciting harmonics and the natural frequency
modes of the blade which are the critical conditions causing the cracking problems. It was then concluded that the dynamic
stress distribution in conjunction with the magnitude of the maximum static stress on the blade, resulting from the centrif-
ugal force, causes HCF to occur in the blade [6].
Rama Rao and Dutta [7] presented a new on-condition monitoring approach of the GT’s rotor based on signal analysis of
the casing vibration (outer casing instrumented with standard accelerometer). The response of the casing due to the excita-
tion by the blade passing frequencies (BPF) of different stages is analyzed to diagnose the condition of the blades. During any
load drop or transient regime of the turbine, the vibration amplitude peaks at the harmonic rotor frequencies are interacted
with blades’ natural frequencies causing disturbance of the functional regime and inducing high vibration and noise in the
entire engine. When a rotating blades vibrate, the amplitude of the BPF show significant variation. Therefore looking at the
waterfall graph, any unusual side bands close to the BPF excitation peaks is indicating abnormal dynamic loading on the
blade such axial loading by pulsating air and this could indicate crack in the blades. In some cases, significant unbalance
of the machine component reveals also second side band on either sides of BPF. Investigations revealed that the proactive
finding of the root cracks identified in five blades of rotor stage #2 (R2) of 210 MWe GT were based on appearance of side
band to the BPF of blades in rotor stage R2.
FOD (Foreign Object Damage) is also a failure’s root cause of the first row of compressor blades [10,11]. In fact, the inlet
duct of the GT engine ingests large amounts of air sucked-in by the compressor during the GT operation period. Therefore,
any solid material entrained with the air will cause damage through either erosion or impact. Nevertheless, FOD is infre-
quent for the industrial GT contrarily to the aviation GT engines. In fact, industrial GTs are equipped with highly efficient
air filtration systems for operation in a variety of environments. The available technologies (static, self cleaning, etc.) provide
barrier filtration against dust, pollen, dirt and other airborne particulate with more than 99% efficiency on sub-micron
particles.
Other studies revealed that the environment must be considered when evaluating any blade failure. Environmental prob-
lems do not normally result in catastrophic failures, but work in conjunction with other failure modes leading to the com-
pound premature failure. Compressor blades (with uncoated airfoils) are frequently affected by corrosion and pitting, that
can be severe if the ambient air contains salts or other contaminants [12–14]. Thus, the blade fatigue strength is significantly
reduced by corrosion in which the stress versus number of cycles (S–N) curve loses its validity and blade failures caused by
crevice corrosion will show symptoms typical of stress corrosion fatigue or stress corrosion.
Ziegler et al. [13] investigated a turbine blade failure of a power plant. Their analysis revealed that damage initiated by
the presence of pits found on the tail of the blade profile caused by corrosive atmosphere, mainly sodium and chloride salts.
The formation of pits lead to residual stresses which give rise to the initiation of intergranular micro-cracks type. Bhagi et al.
[14] found that the fractographic investigation of a failed blade was also initiated by corrosion pits. In fact, the oxides of sil-
icate and sodium were detected on the fractured surface which leads to formation of corrosion pits/attacks. The transgran-
ular cleavage fracture mode and the fatigue marks show that the cause of failure is corrosion fatigue. For both case studies,
examination of the fractured surfaces by energy dispersive spectrometry X-ray (EDS) revealed ample amounts of nonmetallic
inclusions. The solubility of these foreign particles in the air, decreases but condense onto the surface of blades at higher
concentration enhancing the corrosion phenomena.
The above studies and investigations provide an overview of the probable root causes of the turbine blade’s premature
failures, and offer support to analyze the case study subject to this work.
This paper starts by presenting the occured incident and subsequent damages (Section 2). Section 3 focus on the blade
material’s chemical composition, mechanical properties, and required manufacturing process. Section 4 is devoted to the
failure identification: Visual Inspection, Fluorescent Penetrant Inspection (FPI), Stereo-Microscopy, Metallography, Chemical
Analysis, Scanning Electron Microscopy (SEM), Energy Dispersive X-ray Spectroscopy (EDS), and Hardness Testing. A FE anal-
ysis will then be performed on the blade geometry to identify the stress concentration areas, stress/strain values, vibration
modes, . . .and compare results with the fractured material. The last section aims at providing the most probable cause of
failure and recommendations to mitigate such incidents.

2. Failure background

A failure of a first stage compressor blade of a GTG in a Tunisian Gas Treatment plant operated by BG Tunisia caused
severe mechanical damage to the compressor section and power supply troubles. The subject engine was operating in an
island load share capacity where two engines, units A and B, were operating with approximately 30% load. The failure of unit
B occurred immediately following a 60% load increase when unit A suddenly shutdown. The failed engine accumulated only
(5996) hours of service but (102) starts since new installation (Fig. 1).
 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101 91

Fig. 1. Compressor rotor assembly showing fractured blades: (a) overall view of the damaged rotor assembly and (b) overall view of the failed 1st stage
compressor blades.

According to API616 ‘‘Gas Turbines for Petroleum, Chemical and Gas Industry Services’’, the engine shall be designed and
constructed for a minimum service life of 20 years and at least 3 years of interrupted operation. The required time between
inspections shall be not less than 8000 operating hours. The tips of rotating blades and the labyrinths of shrouded rotating
blades shall be designed to allow the unit to start up at any time in accordance with the vendor requirements. Blades shall be
designed to with-stand operation at resonant frequencies during normal warm-up and must have at least 8000 trouble-free
operating hours on similar operating conditions. However, the association of service induced micro structural degradation to
the changes in mechanical properties leads to blade’s material creep properties reduction and premature failures.

3. Metallurgical characterization

The failed blades were made of Inconel 718 which is a Ni–Cr–Fe–Nb alloy fulfilling good mechanical properties such as
high tensile and yield strength, significant rupture strengths and good resistance to mechanical fatigue.

3.1. Chemical composition and mechanical properties of blade material: Inconel 718

Chemical composition of the material UNS N07718 (Formerly Grade 718) complying with ASTM B637 requirements is
prescribed in Table 1.
ASTM B637 Alloy 718 product is available in forged bar, blank, ring, and rolled bar. The material is heat treated by solution
and precipitation hardening. The recommended heat treatment as specified in the ASTM standard is solution treatment at a
temperature of 924–1010  C (1700–1850  F), hold for at least half hour and then cooled down at rate equivalent to air cool or
faster. This heat treatment is to be followed by precipitation hardening treatment at a temperature of 718 ± 14  C
(1325 ± 25  F), hold at temperature for 8 h, cool down to 621 ± 14  C (1150 ± 25  F), and hold until total precipitation heat
treatment time has reached 18 h then air cooled down to room temperature. Inconel 718 alloy differentiates from other
Nickel based super-alloys with the relatively high contents of iron [Fe-19%] and Niobium (or Columbium) [Nb-5%].
Chrome element Cr offers a corrosion resistance properties by forming a layer of oxide Cr2O3 protecting the surface of the
alloy. In order to maintain the forging capabilities of the alloy, Cr elements should not exceed 19% of the material chemical
composition [15].

Table 1
Chemical composition of UNS N07718 – ASTMB637.

Element Composition limits (%) Product (check) analysis variations, under min or
over max, of the specified limit of element
UNS N07718 (Formerly Grade 718)
Carbon 0.08 max 0.01
Manganese 0.35 max 0.03
Silicon 0.35 max 0.03
Phosphorus 0.015 max 0.005
Chromium 17.0–21.0 0.25
Cobalt 1.0 max 0.03
Molybdenum 2.8–3.3 0.05 under min, 0.10 over max
Columbium (Nb) + tantalum 4.75–5.50 0.15 under min, 0.20 over max
Titanium 0.65–1.15 0.04 under min, 0.05 over max
Aluminium 0.20–0.80 0.05 under min, 0.10 over max
Boron 0.006 max 0.002
Iron Remainder ...
Copper 0.30 max 0.03
Nickel 50.0–55.0 0.35
92 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101

Molybdenum element Mo improves the mechanical properties of the alloy at high temperatures due to the solid solution
hardening. Similar to the Chrome element, higher percentage of Mo could affect the forging capabilities of the material. Tita-
nium element Ti used for precipitation of c0 phase should not also exceed 1% of the composition in order to provide the opti-
mum tensile and creep properties. It was also shown that increasing the mass fraction of the Niobium Nb, as essential
element of the d (Ni3Nb) phase, improves the hardness and the yield strength of the material [16].
Fu et al. [17] works state that optimum mechanical properties of the alloy 718 are with a chemical composition including
1% of aluminum, 1% of titan and 5.5% of Niobium.
Based on the above specified requirements, the chemical composition of the fractured blade will be compared with the
ASTM B637 tolerances and the recommended studies’ thresholds which aim to ensure the optimum blade mechanical
characteristics.

3.2. Precipitation of carbides

The metallurgical structure of Inconel 718 is complex due to the precipitation of phases c0 ; c00 and d in the main austenitic
structure of the c phase. The c0 ; c00 and d phases are an ordered structure however the c phase is a disordered lattice. In this
section, we are focusing on the carbides precipitation process in the grain boundaries which could lead to material fracture if
improper morphology appears during the fabrication process.
Carbon element precipitates with Chrome, Titane and Niobium to form primary carbide of type MX (M = Nb, Ti, . . .) and
(X = C, N) such NbC or TiC located in grain boundaries. It is a B1 cubic stable structure at high temperature and it is not
affected by post heat treatment (up to 1250  C). Secondary carbides of type M23C6, M6C or M7C3 [18] could precipitates in
grain boundaries but for Inconel 718, the precipitation of these elements with very low ratio/quantities depends on the con-
centration of Mo and Cr. Carbide can form within the grains and along the grain boundaries.
Therefore, manufacturing process of Inconel 718 turbine blade (raw material’s chemical composition, special heat treat-
ment, casting, forging, etc.) should be well monitored and controlled to avoid chemical or structure anomalies, residual stres-
ses and surface defects in the form of simulators of dents affecting the material strength and leading to premature fracture or
crack initiation. Several investigations of cracked GT blades concluded that the failure root cause was initiated or aggravated
by manufacturing anomalies [19–21].

4. Failure identification

Disassembly of the GT engine at the manufacturer’s workshop revealed that a first stage compressor blade had failed at a
location slightly above the blade root area. The compressor rotor and case assemblies were subjected to Visual Inspection,
Fluorescent Penetrant Inspection (FPI), Stereo-Microscopy, Metallography, Chemical Analysis, Scanning Electron Microscopy
(SEM), Energy Dispersive X-ray Spectroscopy (EDS), and Hardness Testing.

4.1. Visual and stereoscopic examination of the compressor case assembly

Relatively clean condition was noticed on the internal and external surfaces of the compressor case assembly. However
the majority of the inlet variable guide vanes (IGV) had fractured at the airfoil roots with significant plastic deformation and
impact damage on the airfoils of the remaining vanes. The overall view of several IGV’s actuating arms revealed significant
distortion but no signs of scratches or drag markings. Detailed view of the IGV bores, actuated arms’ contact surfaces, wash-
ers and bushings revealed no signs of corrosion attack.
Stereo-microscopic view of a fractured IGV presented significant plastic deformation and fractographic features on the
separated airfoil which indicates that the fracture of the IGV was caused by an overload mechanism, apparently due to sec-
ondary impact damage. On the remaining IGV’s the stereo-microscopic view of the airfoil inner radius revealed no signs of
cracking (Fig. 2).

Fig. 2. Stereo-microscopic views: (a) fractured IGV and (b) non damaged IGV.
 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101 93

From the above investigation of the compressor casing no evidence of seizure, corrosion attack, or fatigue crack initiation
was observed. The historical operation logs from the failed GTG were examined and did not record high IGV’s actuator forc-
ing alarms or shutdown.

4.2. Visual and stereoscopic examination of the rotor assembly

The overall status of the compressor rotor assembly revealed a relatively clean condition. The first and third blades of the
stage 1 have been fractured at the airfoil roots, remaining blades present a severe impact damage. The forward tab on the
retainer clip for blade 1 was undamaged while the retainer clips for all the other 1st stage blades displayed severely distorted
forward tabs (Fig. 3).
Detailed view of the failed first stage compressor blade #1 revealed fracture features indicating a primary fracture surface
from the airfoil leading edge to the midchord. Investigating the blade #3 of the same stage shows a rough surface morphol-
ogy over the entire fracture from the airfoil leading edge to the trailing edge. All the other rotor’s blades were still tied up to
the rotor dovetail but severe impact damage to the airfoils was noticed. This damage is caused by the crash with the loosen
blades #1 and #3 at high rotating speed.
Stereo-microscopic views of the fracture surfaces for both blade #1 and #3 were performed and compared. According to
the leading-edge fracture surface on the failed 1st stage compressor blade #1, an internal anomaly displaying a golden
coloration was detected. It is a rough surface morphology with a well-defined boundary surrounding the fracture surface.

Fig. 3. Failed 1st stage compressor blade 1 (large red arrow) and 3 (small red arrow). Undamaged retainer clip for blade 1 (black arrow), Retainer clips for all
the other 1st stage severely distorted (white arrows). (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

Fig. 4. Leading edge fracture surfaces of blade #1: (a) morphology view of the anomaly region and (b) pressure side view.

Fig. 5. Midchord fracture surfaces: (a) blade #1 and (b) blade #3.
94 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101

Fig. 6. Trailing edge fracture surfaces: (a) blade 1 and (b) blade 3.

The length and width of this region were approximately 5.08 mm and 1.016 mm, respectively (Fig. 4a). Looking at the
pressure-side of the leading-edge area revealed that the anomaly and surrounding fracture surface were approximately
2.54 mm above the blade dovetail (Fig. 4b). The above described anomaly was not detected at the stereoscopic views
of blade #3.
The fracture surface of the midchord area for blade #1 was relatively flat and displayed well-defined crack growth marks,
while the trailing-edge fracture surface presented a rough surface morphology typical of a final overload region. The overload
region accounted for approximately 50% of the airfoil cross-section (Fig. 5).
On the other hand, the leading-edge and trailing-edge fracture surfaces of blade #3 revealed a rough surface morphology.
Contrarily to blade #3 finding, the midchord fracture surface presented significant plastic deformation but no crack growth
marks (Fig. 6).
In addition, the examination of the dovetail surface on the failed 1st stage compressor blade #1 presented only a uniform
loading marks.

4.3. Scanning electron microscopy examination of the failed blades

A scanning electron microscopy (SEM) examination has been performed on the fracture surfaces of the failed blades #1
and #3 of the compressor rotor’s first stage. Low-magnification SEM view of the leading-edge area of the blade #1 displayed
a well-defined boundary of the anomaly region. This anomaly area shows a fractographic features different from the adjacent

Fig. 7. SEM view of blade 1# leading-edge fracture surface: (a) anomaly region (arrows) and (b) crack growth marks (arrows) at an area adjacent to the
anomaly region (bottom).

Fig. 8. High-magnification SEM view of blade 1# leading-edge and midchord fracture surfaces: (a) anomaly region revealing morphological features of
intergranular separation and (b) midchord fracture surface revealing morphological features of transgranular separation.
 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101 95

Fig. 9. SEM view of blade 1# leading-edge: (a) crack growth marks (arrows) at an area adjacent to the anomaly region and (b) scale formation at an area
adjacent to the anomaly region.

Fig. 10. High-magnification SEM view of the midchord fracture surface.

Fig. 11. SEM view of the trailing-edge fracture surface: (a) failed blade #1 and (b) failed blade #3.

area. It also shows a significant plastic deformation and evidence of ductile tearing (Fig. 7a). The adjacent area displayed a
crack growth marks (Fig. 7b).
High-magnification SEM view of the leading-edge anomaly region revealed morphological features of intergranular sep-
aration. However similar view of the midchord area displayed limited plastic deformation and morphological features of
transgranular separation (Fig. 8).
Careful SEM view of the leading-edge fracture surface on failed blade #1 revealed crack growth marks (arrows) at an area
adjacent to the anomaly region (Fig. 9a). A scale formation appears also at an area adjacent to the anomaly region (Fig. 9b).
High-magnification SEM view of the midchord fracture surface displayed also a fine fatigue striations characteristic of a
HCF crack propagation mechanism. The arrows displayed on (Fig. 10) indicate the fatigue crack propagation direction.
SEM view of both the trailing-edge fracture surface of blade #1 and failed blade #3 revealed ductile dimples typical of a
final tensile overload region. No signs of oxidation was found (Fig. 11).

4.4. Energy Dispersive Spectroscopy (EDS) examination

A comparative EDS examination has been performed at the anomaly area and at a second area away from the anomaly on
the fracture surface of the failed blade #1. The carbon content in the anomaly region was found significantly higher (Fig. 12).
The EDS spectrum of the dark region noticed on the leading-edge’s fracture surface of the blade #1 displayed a compo-
sition consistent with Niobium carbide (Fig. 13). Therefore, we suspect that the billet used for the blade forging contains
localized carbon-rich area that originated during casting of the billet.
96 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101

Fig. 12. EDS spectrum of the anomaly region (red line) and an area away from the anomaly (Black line)-Blade #1. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 13. EDS spectrum of the leading-edge dark region – blade #1.

Apart of the high carbon and Niobium contents (C > 0.3%; Nb > 6%; Iron < 18.1%) beside the Niobium carbide along the
grain boundaries of the anomaly region, the failed and intact first stage compressor blades displayed satisfactory chemistry
in compliance with requirements of Table 1. The microstructure of the failed blade in areas away from the anomaly region
was similar to those of the intact blades and was in compliance with specifications.
A hardness test has been performed on the failed first stage compressor blade at the anomaly region and value was the
same as that at areas away from the anomaly region. All hardness values met the hardness requirement for compressor
blades.

4.5. Magnification micrograph examination of radial cross-sections from blade #1

A radial metallographic cross-sections were prepared from the leading-edge and midchord regions of the failed blade #1
(Fig. 14).
A magnification micrograph examination of the leading edge cross-section 1L revealed a dark region with a depth of
approximately 6 lm from the surface with the anomaly. The dark region presents a large amount of precipitates along
the grain boundaries with irregular shapes and forming continuous networks at some areas (Fig. 15a). Similar examination
of the area away from the dark region revealed a limited amount of precipitates along the grain boundaries (Fig. 15b). A forg-
ing lap was also identified at the transition radius of the cross-section 1L under the blade platform with signs of microstruc-
tural changes and shear bands (Fig. 15c).
Micrograph of a cross-section 1M through the midchord fracture surface of blade #1 revealed limited plastic deformation
and a transgranular fracture path consistent with an HCF mechanism. The microstructure was typical of a heat-treated alloy
 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101 97

Fig. 14. Location of radial cross-sections from blade #1.

Fig. 15. Micrograph of a leading-edge cross-section of blade #1: (a) microstructure of the anomaly region; (b) microstructure of the area away from the
anomaly region; and (c) location of forging lap.

Fig. 16. Microstructure of failed blade #1 at the airfoil midchord.

forging, consisted of smaller recrystallized grains with a size of ASTM #9, and larger deformed grains with a size of ASTM #5
(Fig. 16).

5. FE analysis

The stresses acting on the blade in the steady state condition are a combination of aerodynamic loads and centrifugal
loads. The mechanical properties of Inconel 718 at 350  C are considered. The modulus of Young E = 199 GPa, the Poisson’s
ratio m = 0.3 and yield strength ry = 1034 MPa.

5.1. Centrifugal forces

The stress due to centrifugal load is the most critical load acting on the blade and depends on two variable parameter: the
whirling rotor speed and the distance of each element from the rotating axis [6].
The normal stress associated with this force rc is calculated as following [22]:

MV 2 4p2 Mr c w2
rc ¼ ¼ ð1Þ
Ar c A

where M is the mass of the airfoil, V the surface velocity, r c the radius of the mass center from the rotation axis, A the area of
the cross section, and w is the rotor speed in rps.
98 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101

5.2. Aerodynamic loads

Assessment of aerodynamic loads in a turbine stage (Stator/Rotor) is complex and depends on the blade/vane passage
geometry such the twisting angle along the span, the end wall contouring at the Hub/Root level, and the clearance between
the blade tip and the shroud [23]. Therefore, the passage flow is characterized by the boundary layer effects, the secondary
flows generated by the passage pressure gradients, and the vortical flow such as: the leading edge ‘‘horse-shoe’’ vortices, tip-
leakage flow vortices and corner vortices [24].
For the present case study, the minor loads due to the secondary flows are neglected comparing to the other efforts. Fluid
dynamics are considered only for the blade’s mid span region located away from the Hub/Root endwall region and the blade
Tip/Shroud clearance. In fact, the pressure distribution does not change along most of the blade span or height except near
the hub or tip region. The flow is then assumed two dimensional and radial flow is negligible (Fig. 17).
Graphs of Fig. 18 show the pressure and velocity distribution in the mid-span plane along a typical blade passage. Mea-
surements and computation of the flow mathematical model have been used to estimate the distribution of the static pres-
sure coefficient C p , which is determined from the difference of blade surface pressure and reference pressure at the passage
inlet dynamic pressure [25,26].

P  Pref
Cp ¼ ð2Þ
0:5qV 2
The aerodynamic loading of the blade is determined based on the area subject to such pressure curves as shown in Fig. 18.
The highest pressure coefficient C p is at the stagnation point located at the blade pressure surface close to the leading edge.
The low value of C p is located on the blade suction surface at the passage throat area (40–50% of the axial chord). On the
blade pressure side, the velocity increases from the leading edge toward trailing edge contrarily to the C p value. On the blade
suction side, the velocity increases to the throat location and declines when it encontours the adverse pressure gradients
(increased C p ).

5.3. FE analysis of the blade

The compressor first stage blade is simulated by the FE method using the software Abaqus. C3D10 quadratic tetrahedral
elements were used for the mesh generation (Fig. 19). The mesh consists of (26,007) elements and (42,093) nodes. The
applied mechanical properties of the blade material were the Inconel 718 at 350  C (Modulus of elasticity, Poisson’s ratio
and density). In order to determine the stress regions and to simplify the static analysis, an elastic behavior of the material
has been considered.
The applied boundary conditions were centrifugal and aerodynamic loads as described in Sections 5.1 and 5.2. Further-
more, the fixed (Anchor) part was utilized on the blade root region. Fig. 20 shows the result of the static analysis at the
blade’s section and pressure surfaces. The maximum value of von Mises stress is about 513 MPa, which in comparison of
the yield strength of the material 1034 MPa is within safe limits.
A high stress region is noticed on the blade suction surface due to the effect of centrifugal and aerodynamic pressure
distribution (Blade twist angle), which induced torsional moment along the blade length and especially close to the blade
root where the crack was initiated in the failed blade. At midchord blade region close to the blade’s root, another stress

Fig. 17. Streamlines distribution in the mid-span plane along blade passage [25].
 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101 99

0.5

-0.5
Blade pressure surface
Blade suction surface
-1

-1.5

-2
Leading edge Trailing edge
-2.5
0 10 20 30 40 50 60 70 80 90 100
Cp vs %axial chord
(a)

1.4

1.2

0.8
Blade pressure surface
Blade suction surface
0.6

0.4

0.2
Leading edge Trailing edge
0
0 20 40 60 80 100
V/Vref vs %axial chord
(b)

Fig. 18. Pressure and velocity distribution on typical blade surface: (a) pressure distribution and (b) velocity distribution [25].

Fig. 19. Mesh model of the blade.

concentration area with von Mises stress value about 342 MPa is explaining the observed typical ductile dimples due to final
tensile overload region as presented in Section 4.3. The maximum value of von Mises stress value at the blade trailing edge is
due to the reduced blade section area. The FE analysis is showing that the crack should be initiated at the blade leading edge
from the suction surface toward the pressure surface.
It is worth noting that both ultrasonic and rotary-bending fatigue testing (Re ¼ 1) performed on Inconel 718 specimens
by Zhang et al. [27] confirmed that fatigue failure of this superalloy continues to occur after exceeding 107 cycles (conven-
tional fatigue limit) at a stress amplitude above 540 MPa which is exceeding the maximum stress applied to the blade airfoil.
100 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101

Fig. 20. FE analysis of the first stage blade: (a) blade pressure side and (b) blade suction side.

6. Failure analysis and recommendations

Examination of the failed 1st stage compressor blade #1 revealed cracks initiated from the anomaly region and propa-
gated towards airfoil midchord under moderate load levels until final tensile overload separation occurred. Boroscopic
inspection records did not show any initiation of corrosion or pitting marks on the first stage blades. The service environ-
ment is not considered a corrosive environment. Looking at the inlet duct filters, the status was good and induced FOD is
discarded from the root cause analysis.
Deep investigation of the anomaly region displayed morphological features of intergranular separation, and a golden col-
oration indicating that the fracture at this region was produced by tensile overload at temperature above 1000  F, most likely
during the forging process. The microstructure at the anomaly region showed a much higher density of Niobium carbide pre-
cipitates. This suggested that the billet used for the blade forging contained a localized carbon-rich area that originated dur-
ing casting of the billet, which led to the formation of large amounts of brittle carbides along the grain boundaries and
subsequently resulted in internal grain-boundary cracking during forging.
A forging lap was also identified at the blade #1 leading edge. It’s located at the transition radius under the blade platform
where shear bands and signs of microstructural changes due to inhomogeneous plastic deformation were observed.
No foreign materials were identified inside the forging lap. Apart of the high carbon and Niobium contents (C > 0.3%;
Nb > 6%; Iron < 18.1%) localized at the anomaly region, the failed and intact first stage compressor blades displayed satisfac-
tory chemistry in compliance with ASTM B637 requirements. The microstructure of the failed blade in areas away from the
anomaly region was similar to those of the intact blades.
The airfoil fracture of the other 1st stage compressor blade #3 was produced by a tensile overload mechanism due
to impact of the fragment from the primarily failed blade #1. No fatigue cracks were identified in the other 1st stage
compressor blades or in any blade of the other compressor stages of the subject unit.
 W. Maktouf, K. Saï / Engineering Failure Analysis 47 (2015) 89–101 101

7. Conclusion and recommendations

Based on the analysis, the failure of the 1st stage compressor blade from the subject GT occurred by a high-cycle fatigue
(HCF) mechanism. The root cause of the failure was attributed to an internal metallurgical anomaly near the airfoil leading
edge. Fatigue cracks initiated from the anomaly region and propagated towards airfoil mid-chord until final tensile overload
separation occurred.
Therefore, manufacturing process of the GT blades should be well monitored and controlled to avoid residual stresses or
surface defects. Random checks on blades from each forging heat lot will reduce such risks. Operator should implement a
rigorous on-condition monitoring of the GT rotor and spot any side bands close to the BPF excitation peaks. Load drop or
transient regime of the turbines increases vibration amplitude when harmonic rotor frequencies are interacted with blade’s
natural frequencies. Thus, the number of engine’s start-up and shutdown of the GT shall be reduced to the minimum. Rou-
tine boroscopic inspection of the GT rotor should focus on the stress concentration area of the blade located near the con-
nection region of the airfoil and the root as shown by the FE analysis.
The static loading analysis performed in this work needs to be completed by dynamic analysis to assess the blade resis-
tance to fatigue induced by repeated/fluctuated loads and the aerodynamic cyclic stresses. A thermo-elastoplastic behavior
of the blade material is required for identifying the blade’s strain-life fatigue crack initiation and propagation.

Acknowledgment

The authors wish to thank the BG-Group Rotating Equipment Technical Authority for the technical support and for
encouraging any advanced technical investigation on failures of the rotating equipments in BG assets.

References

[1] Carter T. Common failures in gas turbine blades. Eng Fail Anal 2005;12:237–47.
[2] Meher-Homji B, Gabriles G. Gas turbine blade failures – causes avoidance and troubleshooting. In: Proceedings of the 27th turbomachinery
symposium, vol. 27; 1995. p. 129–80.
[3] Dundas R. Engineering and metallographic aspects of gas turbine engine failure investigation: identifying the causes. Flight Saf Found – Aviat Mech Bull
1994:1–20.
[4] Kargarnejad S, Djavanroodi F. Failure assessment of Nimonic 80 A gas turbine blade. Eng Fail Anal 2012;26:211–9.
[5] Qu S, Fu C, Dong C, Tian J, Zhang Z. Failure analysis of the 1st stage blades in gas turbine engine. Eng Fail Anal 2013;32:292–303.
[6] Poursaeidi E, Babaei A, Mohammadi Arhani M, Arablu M. Effects of natural frequencies on the failure of R1 compressor blades. Eng Fail Anal
2012;25:304–15.
[7] Rama Rao A, Dutta B. Vibration analysis for detecting failure of compressor blade. Eng Fail Anal 2012;25:211–8.
[8] Witek L. Experimental crack propagation and failure analysis of the first stage compressor blade subjected to vibration. Eng Fail Anal 2009;16:2163–70.
[9] Mazur Z, Garcia-Illescas R, Porcayo-Calderon J. Last stage blades failure analysis of a 28 MW geothermal turbine. Eng Fail Anal 2009;16:1020–32.
[10] Witek L. Numerical stress and crack initiation analysis of the compressor blades after foreign object damage subjected to high-cycle fatigue. Eng Fail
Anal 2011;18:2111–25.
[11] Silveira E, Atxaga G, Irisarri A. Failure analysis of a set of compressor blades. Eng Fail Anal 2008;15:666–74.
[12] Lourenço N, Graça M, Franco LAL, Silva O. Fatigue failure of a compressor blade. Eng Fail Anal 2008;15:1150–4.
[13] Ziegler D, Puccinelli M, Bergallo B, Picasso A. Investigation of turbine blade failure in a thermal power plant. Case Stud Eng Fail Anal 2013;1:192–9.
[14] Bhagi L, Gupta P, Rastogi V. Fractographic investigations of the failure of L-1 low pressure steam turbine blade. Case Stud Eng Fail Anal 2013;1:72–8.
[15] Niang A. Contribution à l’étude de la précipitation des phases intermetalliques dans l’alliage 718 (in French). PhD thesis, INPT Toulouse, France; 2010.
[16] Rizzo J, Buzzanell J. Effect of chemistry variations on the structural stability of alloy 718. J Met 1969;21:501–43.
[17] Fu S, Dong J, Zhang M, Xie X. Alloy design and development of Inconel 718 type alloy. Mater Sci Eng 2009;499:215–20.
[18] Sabol G, Stickler R. Microstructure of nickel-based superalloys. Phys Status Solidi 1969;35:11–52.
[19] Zheng L, Xiao C, Zhang G. Brittle fracture of gas turbine blade caused by the formation of primary -nial phase in ni-base superalloy. Eng Fail Anal
2012;26:318–24.
[20] Troshchenko V, Prokopenko A. Fatigue strength of gas turbine compressor blades. Eng Fail Anal 2000;7:209–20.
[21] Perkins K, Bache M. The influence of inclusions on the fatigue performance of a low pressure turbine blade steel. Int J Fatigue 2005;27:610–6.
[22] Mukhopadhyay N, Ghosh Chowdhury S, Das G, Chattoraj I, Das S, Bhattacharya D. An investigation of the failure of low pressure steam turbine blades.
Eng Fail Anal 1998;5:181–93.
[23] Lakshminarayana B. Fluid mechanics and heat transfer of turbomachinery. New York: John Wiley and Sons Inc.; 1996.
[24] Dixon S. Fluid mechanics, thermodynamics of turbomachinery. 3rd ed. Oxford: Butterworth-Heinemann Ltd; 1995.
[25] Acharya S. Endwall cooling with endwall contouring and leading edge fillet. Semi-annual report submitted to UTSR, South Carolina, Project No. 02-01-
SR098; 2003.
[26] Bohn D, Kusterer K, Sürken N, Kreitmeler F. Influence of endwall contouring in axial gaps on the flow field in a four-stage turbine. In: ASME proc turbo
expo, 2000-GT-472; 2000.
[27] Zhang Y, Duan Z, Shi H. Comparison of the very high cycle fatigue behaviors of INCONEL 718 with different loading frequencies. Sci China – Phys Mech
Astron 2012;56:617–23.